[MPS] Enable bfloat16 testing #136987
Closed
[MPS] Enable bfloat16 testing #136987
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
[ghstack-poisoned]
This was referenced Sep 30, 2024
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/136987
Note: Links to docs will display an error until the docs builds have been completed. ✅ No FailuresAs of commit d3b8428 with merge base dfe1d45 ( This comment was automatically generated by Dr. CI and updates every 15 minutes. |
pytorch-bot
bot
added
ciflow/mps
This was referenced Sep 30, 2024
malfet
added
topic: improvements
[ghstack-poisoned]
This was referenced Sep 30, 2024
pytorchmergebot
pushed a commit
that referenced
this pull request
Sep 30, 2024
Before this change, test failed with unable to compile errors, as `bfloat16` requires explicit cast. Tested in #136987 Pull Request resolved: #136981 Approved by: https://github.com/Skylion007
pytorchmergebot
pushed a commit
that referenced
this pull request
Sep 30, 2024
) Just adds instantiation of the kernels and sometimes explicit cast. Tested in #136987 Pull Request resolved: #136982 Approved by: https://github.com/Skylion007 ghstack dependencies: #136981
pytorchmergebot
pushed a commit
that referenced
this pull request
Sep 30, 2024
By simply adding explicit instantiation Tested in #136987 Pull Request resolved: #136983 Approved by: https://github.com/Skylion007 ghstack dependencies: #136981, #136982
[ghstack-poisoned]
[ghstack-poisoned]
By even further reducing precisions of imprecise FP16 ops, introducing new BF16_LOW_PRECISION_OPS category and marking BF16 tests as xfail for `divfloor_rounding`, `floor_divide` and `remainder` [ghstack-poisoned]
malfet
added a commit
that referenced
this pull request
Oct 1, 2024
For Metal cast ops to comple, one need to explicitly cast to/from `bfloat` unlike for other dtypes Tested in #136987 [ghstack-poisoned]
albanD
approved these changes
Oct 1, 2024
Collaborator
albanD
left a comment
albanD
left a comment
There was a problem hiding this comment.
Any chance we can fix that behavior? This will most likely change the numerics for end users significantly.
@kulinseth ?
pytorchmergebot
pushed a commit
that referenced
this pull request
Oct 1, 2024
For Metal cast ops to comple, one need to explicitly cast to/from `bfloat` unlike for other dtypes Tested in #136987 Pull Request resolved: #137070 Approved by: https://github.com/Skylion007
By even further reducing precisions of imprecise FP16 ops, introducing new BF16_LOW_PRECISION_OPS category and marking BF16 tests as xfail for `divfloor_rounding`, `floor_divide` and `remainder`. I guess the nature of low-precision results, is that MPSGraph, unlike the rest of the PyTorch does not do accumulation over fp32 for reduction operations [ghstack-poisoned]
Contributor
Author
|
@pytorchbot merge -f "MPS + Lint are green" |
Collaborator
Merge startedYour change will be merged immediately since you used the force (-f) flag, bypassing any CI checks (ETA: 1-5 minutes). Please use Learn more about merging in the wiki. Questions? Feedback? Please reach out to the PyTorch DevX Team |
AnantGulati
pushed a commit
to AnantGulati/pytorch
that referenced
this pull request
Oct 2, 2024
Before this change, test failed with unable to compile errors, as `bfloat16` requires explicit cast. Tested in pytorch#136987 Pull Request resolved: pytorch#136981 Approved by: https://github.com/Skylion007
AnantGulati
pushed a commit
to AnantGulati/pytorch
that referenced
this pull request
Oct 2, 2024
…rch#136982) Just adds instantiation of the kernels and sometimes explicit cast. Tested in pytorch#136987 Pull Request resolved: pytorch#136982 Approved by: https://github.com/Skylion007 ghstack dependencies: pytorch#136981
AnantGulati
pushed a commit
to AnantGulati/pytorch
that referenced
this pull request
Oct 2, 2024
By simply adding explicit instantiation Tested in pytorch#136987 Pull Request resolved: pytorch#136983 Approved by: https://github.com/Skylion007 ghstack dependencies: pytorch#136981, pytorch#136982
AnantGulati
pushed a commit
to AnantGulati/pytorch
that referenced
this pull request
Oct 2, 2024
By adding explicit instantiation. Tested in pytorch#136987 Pull Request resolved: pytorch#136984 Approved by: https://github.com/Skylion007 ghstack dependencies: pytorch#136981, pytorch#136982, pytorch#136983
AnantGulati
pushed a commit
to AnantGulati/pytorch
that referenced
this pull request
Oct 2, 2024
For Metal cast ops to comple, one need to explicitly cast to/from `bfloat` unlike for other dtypes Tested in pytorch#136987 Pull Request resolved: pytorch#137070 Approved by: https://github.com/Skylion007
AnantGulati
pushed a commit
to AnantGulati/pytorch
that referenced
this pull request
Oct 2, 2024
By even further reducing precisions of imprecise FP16 ops, introducing new BF16_LOW_PRECISION_OPS category and marking BF16 tests as xfail for `divfloor_rounding`, `floor_divide` and `remainder`. I guess the nature of low-precision results, is that MPSGraph, unlike the rest of the PyTorch does not do accumulation over fp32 for reduction operations Pull Request resolved: pytorch#136987 Approved by: https://github.com/albanD ghstack dependencies: pytorch#137070
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Stack from ghstack (oldest at bottom):
By even further reducing precisions of imprecise FP16 ops, introducing new BF16_LOW_PRECISION_OPS category and marking BF16 tests as xfail for
divfloor_rounding,floor_divideandremainder.I guess the nature of low-precision results, is that MPSGraph, unlike the rest of the PyTorch does not do accumulation over fp32 for reduction operations